RFC 1195 [IS-IS for IP Networks] is an interior gateway protocol utilized to determine the topology and routing within a network with the network divided between level 1 areas and a level 2 domain connecting the level 1 areas. IS-IS works in conjunction with a link state protocol that advertises the adjacencies of each of the nodes in an area or domain of an IS-IS network. The IS-IS protocol utilizes a route computation algorithm such as shortest path first (SPF) algorithm to calculate routes from each network device in an IS-IS area or IS-IS domain to each other network device in that area as well as routes to the border routers that enable data to be forward to destinations outside of the IS-IS area or IS-IS domain. Each network device in the IS-IS area or IS-IS domain constructs a routing information base in which the next hop to each other destination network device in the IS-IS area or IS-IS domain is recorded to be utilized to forward data packets.
When an adjacent router or the link to that adjacent router designated as the next hop for a route to a given destination network device fails, traffic disruption will occur until the network re-converges (i.e., recalculates paths to each of the network devices in the IS-IS area or IS-IS domain). This disruption of traffic can be prevented by Loop Free Alternatives (LFA) technology in principle as specified in RFC 5286. LFAs are alternate routes to each network device that can be utilized in the case that a next hop to a given destination is not available.
In many IS-IS areas or IS-IS domains network devices are associated with a set of prefixes, i.e., ranges of Internet Protocol addresses that are reachable from the network device. In some cases, a prefix may be associated with multiple network devices, which is referred to as a ‘multi-homed prefix.’ This increases the complexity of calculating LFAs to these multi-homed prefixes. However, the specification (RFC 5286) doesn't specify any procedure that can be applied to efficiently identify multi-homed prefixes. As a result, in many instances the standard LFA computation process may cause excessive and unnecessary computation for each prefix to have an optimal LFA.